How to Actually Adopt AI in Your SaaS Marketing Department

Most SaaS companies say they’re using AI for marketing. What many mean is someone on the team asks ChatGPT to rewrite a LinkedIn post every now and then.

There’s a big gap between that and an integrated AI marketing operation. Closing it takes more than signing up for a Claude licence. It takes mapping your workflows, choosing the right tools for the right jobs, and building a process that actually runs day to day. At Xander Marketing, we’re a B2B SaaS marketing agency that’s been building AI into our client work since 2022 and this guide is based on what we’ve learned.

Here’s how to do it, in six steps.

1. Audit Your Marketing Operations

Before you think about AI, you need a clear picture of how your marketing time is spent.

Map everything. Recurring tasks, workflows, handoffs between people and tools. Content production, email campaigns, reporting, social scheduling, competitor monitoring, onboarding sequences. Write it all down in whatever format works for you.

Be specific about it. “We do content marketing” tells you nothing. “Every Tuesday, Chris writes a blog post brief, sends it to the freelancer, reviews the draft on Thursday, edits it, uploads it to WordPress, creates three social posts to promote it, and schedules them in Buffer”, tells you exactly where AI could slot in. Even if Chris is you, the exercise is the same.

Once you’ve got that picture, score each workflow on how often it happens, how long it takes, and how much human judgment it requires. You’re looking for things that happen a lot, eat up time, and don’t need heavy creative or strategic input, things like reporting, content repurposing, first drafts, email personalisation. 

As a bonus, you could rank and prioritise them.

2. What else is now possible? 

Now you know where your time goes, go and look at what else AI makes possible. Not just faster versions of what you’re already doing, but work that wasn’t feasible before.

Most people still think of AI as a writing assistant. It does that fine, but that’s a fraction of it. Companies are using AI to build interactive tools that used to need a developer. They’re connecting AI directly to their CRM and asking it to find patterns in lost deals or flag accounts worth re-engaging. They’re running monthly competitive intelligence that would have cost a consultant’s day rate every time. Landing page builds have gone from days to a couple of hours. Case studies from half a day to half an hour.

Look at your top five workflows from the audit, but also think about what you’ve never done because it was too expensive or too slow. That’s often where AI adds the most value. For a detailed look at what’s possible, including interactive tools, business system integrations, and competitive intelligence workflows, see our companion post: What AI Has Made Possible for Marketing Your SaaS in 2026.

3. Pick Your Tools

You’ve got your workflows mapped, and you’ve seen what’s possible. Now you need to decide what you’re actually going to use.

The big general-purpose AI tools right now are ChatGPT, Claude, Gemini, and Copilot (built on top of ChatGPT). At the time of writing, Claude would be our recommendation, but it depends on your needs and current setup. If you already use Google Workspace or Microsoft products heavily, Gemini or Copilot may be better options for you.

Most companies end up using more than one. That’s fine. What you want to avoid is different people using different tools for the same task with no consistency in how they’re set up.

Beyond the general-purpose tools, look at what’s built into the platforms you already pay for. HubSpot has AI across content creation, email personalisation, lead scoring, and campaign optimisation. Your email platform, social scheduler, and CRM have likely added AI features in the last year. Check before you add new subscriptions. Specialist tools for video, image creation, transcription, and SEO analysis can come later once your core workflows are running.

On cost: the general-purpose tools run £15-25/month per user for their standard paid tiers. You’ll get that back in minutes when you compare the output to what a person can produce in the same time.

4. Build Your AI Marketing System

Once you’ve picked your tools, you need to set them up properly. This means more than writing a few prompts. The tools themselves have building blocks designed for exactly this.

Projects. In Claude (and similar features in other tools), a Project is a workspace where you load persistent context. Your brand voice document, your messaging framework, your audience personas, your competitor information. Everything the AI needs to produce on-brand output goes here. Set up a project for each major area of your marketing – content, campaigns, competitor intelligence, and every output from that project will draw from the same context. If you’re on a team or enterprise account, these projects can be shared across your whole marketing operation so everyone’s working from the same foundation.

Skills. A skill is a set of repeatable instructions for a specific task. Content repurposing, for example: you define the input (a published blog post), the outputs you want (LinkedIn post, email snippet, short-form summaries, video script outline), your brand voice rules, and the format for each. Save that as a skill and you can run it every time you publish something new.

A clever way to build a skill is to do the task manually in AI first. Get the output right through conversation, then ask AI to turn what you just did into a repeatable skill. Test it on a different input. Tweak it. Run it again. Each round gets tighter. A skill that took you an hour of back-and-forth to get right the first time should run in minutes once it’s built.

MCPs and connectors. MCP (Model Context Protocol) lets you connect AI tools directly to your business systems, your CRM, analytics, Google Drive, Slack, and email. Instead of exporting data, copying it into a spreadsheet, and then asking AI to analyse it, you connect the system once and query it directly. Ask your CRM which deals have been stuck at the same stage for 30 days. Ask your analytics which landing pages have the highest bounce rate. 

Workflows. With projects, skills, and connectors in place, you can map end-to-end workflows. A competitor monitoring workflow might look like this: structured prompts covering competitors’ messaging, positioning, pricing, content output, and user reviews, run monthly, with the findings pulled into a single report. A human reads the report, interprets what’s changed, and decides what it means for your positioning or content plan.

All of these need testing and improving over time. AI drifts. Prompts that worked well in week one produce slightly different results by week six as you feed in more context or the tools update. Brand voice is the most common casualty. If you’re not checking outputs against your brand context regularly, quality degrades slowly enough that you don’t notice until it’s a problem. Build review into the process from the start.

5. Build, Run, and Improve

Don’t try to build five workflows at once. Pick the one you scored highest, build it, get it running reliably, and move on.

Write things down as you go, which prompts work, where outputs fall short, what the quality checks catch. This becomes your reference for every workflow that follows, and each one gets faster to build.

Where companies come unstuck is the day-to-day. Someone builds something clever, nobody owns it, and within a month it’s gathering dust. Every workflow needs a person responsible for running it and checking what comes out. It needs a QC process that matches the stakes, because the review you give a social post is very different from the review you give a customer email sequence. And when the AI gets something wrong, there needs to be a clear path for who fixes it and how that fix feeds back in.

Set a review cadence, fortnightly to start with, monthly once things settle. Each review should ask whether the workflow is hitting the quality bar, whether it’s saving the time you expected, and what’s degrading.

A few months in, you’ll start spotting opportunities you couldn’t see before. The more workflows you build, the more you can see what else could be built.

6. Train Your People and Know What to Hire For

Once you’ve built your first few workflows, you need to get the people using them up to speed. That means showing them the tools, walking them through the specific workflows you’ve built, and explaining how to give feedback when something isn’t right.

That feedback part matters. AI works best when you treat it like working with a person. You brief it properly, you review its work, you tell it what it got wrong and how to fix it. The better the feedback, the better the output next time. If someone on your team is accepting mediocre AI output because they don’t know how to push back on it, that’s a training problem.

The bigger shift is in how people think about their work. Everyone involved in marketing should be asking two questions as a habit. First: should I be doing this myself, or should AI be doing it? Not everything should be automated, but plenty of tasks are still being done manually out of habit rather than because they need to be. Second: if I’m going to do this again, how can I build a skill so it’s faster next time? Every repeatable task is a candidate. The more your team thinks this way, the more your AI operation grows without you having to plan every piece of it.

Share what’s working. If someone builds a skill that cuts a weekly task from an hour to ten minutes, make sure everyone knows about it. A shared library of prompts, skills, and workflows means the learning compounds rather than staying in one person’s head.

Your hiring profile changes as the operation matures. When you’re looking for your next marketing hire, look for people who’ve actually built AI workflows, not people who’ve used ChatGPT to tidy up copy. Ask candidates to show you something they’ve built, walk you through how it works, and explain what they’d change. Look for people who can describe why a workflow failed and what they did about it, that tells you more about their ability than any list of tools they’ve used.

Build It Yourself, or Get Help

Everything above is something you can do internally if you’ve got the time and inclination.

If you’d rather not figure it out alone, Xander Marketing can build your AI marketing operation and run it as your outsourced marketing department, or build it and hand it over to your team.

Either way, you skip the learning curve.

The easiest way to get started is to book a free 30-minute marketing consultation.